I have a very serious concern about my model being trained by autogluon’s automl.
After training the model for preset – ‘high_quality’, I try to predict data for say one new data point. As per my analysis one of the predicted variables should change its predicted value but it does not. In fact the predicted value does not go less than the least value of the actual dependent variable under any circumstance which is very concerning for me as I believe the predicted value should go down.

this is my data:

Date        A   B   C   Target  D       E
4/10/2018   34  190 5   72  31.075      0.040818654
5/10/2018   34  190 5   72  30.23636364 0.040818654
6/11/2018   34  190 5   72  30.575      0.040818654
7/15/2018   34  185 5   72  30.66       0.040818654
8/26/2018   34  185 5   72  30.6902439  0.040818654
9/30/2018   34  185 5   72  30.60487805 0.040818654
10/19/2018  34  180 5   72  30.6        0.040818654
11/12/2018  34  180 5   72  30.50769231 0.040818654
12/24/2018  34  190 5   72  30.5        0.040818654
1/20/2019   34  190 5   72  30.4        0.040818654
2/9/2019    34  190 5   72  30.4        0.040818654
3/21/2019   34  185 5   72  30.3        0.040818654
4/12/2019   34  185 5   69  30.27142857 0.040818654
5/13/2019   34  185 5   66  30.2        0.040818654
6/3/2019    34  185 5   66  30.26666667 0.040818654
7/26/2019   25  185 5   66  30.11842105 0.040818654
8/10/2019   25  185 5   59  30.04666667 0.040818654
9/27/2019   25  185 5   59  30.31111111 0.040818654
10/7/2019   25  185 5   59  30.27407407 0.040818654
11/17/2019  25  185 5   59  30.12222222 0.040818654
12/6/2019   28  185 5   59  30.05185185 0.040894755
1/17/2020   28  185 5   59  29.6972973  0.041351364
2/21/2020   28  185 5   59  29.31891892 0.041731872
3/7/2020    28  185 5   59  29.24444444 0.041894947
4/10/2020   29  185 5   59  29.11875    0.042264583
5/5/2020    29  170 5   59  29.33076923 0.042536374
6/6/2020    29  170 5   59  29.37894737 0.042884267
7/9/2020    29  175 5   59  29.3        0.043243031
8/4/2020    29  175 5   59  29.40243902 0.043525694
9/13/2020   29  175 5   59  29.5        0.04396056
10/19/2020  29  175 5   59  29.47142857 0.04435194
11/15/2020  29  175 5   59  29.45       0.044645474
12/6/2020   29  175 5   59  29.43333333 0.044873779
1/23/2021   29  175 5   56  29.4        0.045395618
2/4/2021    29  175 5   56  29.4        0.045526078
3/22/2021   29  166 5   56  29.4        0.046026174
4/16/2021   29  166 5   56  29.4        0.046297965
4/25/2021   29  166 5   51  29.4        0.04639581
5/16/2021   29  166 5   51  29.4        0.046624114
6/22/2021   29  157 5   51  29.4        0.047026365
7/5/2021    29  157 5   47  29.4        0.047167697
8/1/2021    28  160 5   47  29.4        0.047486411
9/16/2021   28  160 5   47  29.33076923 0.048091805
10/10/2021  28  160 5   47  29.4        0.048407662
11/10/2021  28  158 5   47  29.4        0.048815645
12/9/2021   28  158 5   47  29.4        0.049197306
1/28/2022   28  158 5   47  29.69918699 0.049855343
2/20/2022   31  156 5   47  29.99837398 0.05015804
3/20/2022   28  156 5   43  30.36260163 0.05052654
4/14/2022   28  155 5   39  30.68780488 0.050855558
5/25/2022   28  155 5   39  30.43790323 0.051395148
7/18/2022   28  142 5   39  28.65241935 0.052105828
8/7/2022    28  142 5   39  27.99112903 0.052369042
9/18/2022   28  142 5   39  26.92727273 0.052921793
10/31/2022  28  142 5   39  27.04871795 0.053487704
11/22/2022  28  142 5   39  27.09583333 0.05377724
12/30/2022  28  146 5.5 39  27.01666667 0.054277348
1/15/2023   28  146 5.5 39  26.94285714 0.05448792
2/1/2023    28  146 5   39  26.82142857 0.054711652
2/15/2023   28  146 5.5 39  26.91       0.054895903
3/3/2023    28  145 5.5 39  27.07       0.055106474
4/9/2023    28  145 5.5 39  27.99230769 0.055593421
5/9/2023    28  145 5.5 39  28.05       0.055988243
6/2/2023    28  145 5   39  28.6952381  0.056304101
6/28/2023   29  130 5.5 39  30.8        0.05664628
7/1/2023    29  130 5.5 39  30.74       0.056685762
7/16/2023   29  130 5.5 39  30.61612903 0.056883173
8/9/2023    29  130 5.5 39  30.24285714 0.05719903
9/25/2023   29  140 5.5 39  30.48888889 0.057817585
9/26/2023   29  140 5.5 39  30.47777778 0.057830745
9/27/2023   29  140 5.5 39  30.46666667 0.057843906
9/28/2023   29  140 5.5 39  30.45555556 0.057857067
10/28/2023  29  105 5.5 39  30.25294118 0.058251889
11/14/2023  29  105 5.5 39  30.34545455 0.058475621
12/6/2023   29  105 5.5 39  30.4        0.058765157
1/24/2024   29  105 5.5 39  30.4        0.058765157

I have trained autogluon’s Tabular predictor to predict the Target variable with experiment type being Regression. I get a very low MAPE. But when I try to use a data point in order to do an inference on the model, the output never goes below 39. In fact even if I make all the inputs 0, the output gives me a value of 46 which is beyond my capability to understand.

Training example:

predictor = TabularPredictor(label='Target', problem_type='regression', eval_metric='mean_absolute_percentage_error').fit(train_df, presets='high_quality', time_limit=300)

Inference sample:

new_data_point = {
    'A': [20],
    'B': [90],
    'C': [5.0],
    'D': [20],
    'E': [0.004]
}
new_data_df = pd.DataFrame(new_data_point)

predictor.predict(new_data_df) gives output:
0    51.603809
Name: Oil Rate, dtype: float32

this output hardly ever changes no matter how much I change the data point for inference.

Why if the output of the predictor.predict(new_data_df) not changing even though I keep changing the new_data_df values?